A Novel Data-Driven Current Control Design for Grid-Connected Inverter with LCL Filter

This paper presents a novel current control approach for grid-connected inverters employing deep symbolic regression (DSR). The proposed control approach overcomes the limitations of traditional proportional-integral (PI) controllers. Although PI controllers are commonly used due to their simple structure, their practical implementation in real-world grid-connected inverter applications presents computational challenges. Precise gain tuning, real-time integral computation, and additional anti-windup mechanisms significantly increase execution time, particularly in high-speed digital control systems, making them less efficient for applications requiring fast dynamic response and reduced computational complexity. By replacing the conventional PI controller with DSR-based current control, we significantly reduce computational burden and improve adaptability under varied operating conditions. In addition, the proposed controller addresses the decoupling inaccuracies commonly encountered in the conventional PI-based control approach. DSR works offline to provide numerical analytical expressions that give optimal control performance. The simulation results demonstrate that the DSR-based current control gives better steady-state performance and lower harmonic distortion compared to the conventional PI current control. The test results validate the effectiveness of the proposed control scheme, highlighting its potential to improve power quality and operational robustness in practical grid applications, making it a promising alternative for efficient grid-connected inverter control.